基于内容的视频高效检索技术的研究

【作者】 彭希君；

【导师】 刘凤玉；

【作者基本信息】 南京理工大学 ， 计算机应用， 2004， 硕士

【摘要】 近年来，随着计算机和网络通信等技术的高速发展，Internet上的信息种类和数量与日俱增。如何从这个海量的多媒体信息库中找到需要的多媒体信息已经成为一个重要的课题。针对这一需求，本文将研究重点放在最常见的媒质—视频，视频结构化和视频特征的选择组织是视频信息检索中的两个主要技术，本文围绕这两方面展开。 文章首先介绍视频结构化的内容以及相应的各种算法，着重对镜头分割中的自动门限算法进行改进，同时分析经典的关键帧提取方法在MPEG格式的压缩视频中提取关键帧的应用，并详细介绍了MPEG-7的全部描述符及描述方案，研究其在基于内容的特征组织中的应用，并作了相应的实现。 在上述研究的基础上文章给出了通用的基于内容的视频高效检索试验原型系统的设计方案。 最后对本论文的研究内容作了简单的总结，指出视频检索的研究前景和方向。更多还原

【Abstract】 Recently, with the rapid development of the computer, network and communication technologies, the amount and types of information are booming over the Internet day and day. How to get the required information from this huge multimedia library is becoming a very important subject. This thesis is proposed according to this requirement and put emphasis on video. The researching content is about video structure analysis and the description principle of features.Firstly, introductions to the content of video structure analysis and the algorisms about it. According to the importance of shot segmentation in content-based retrieval or CBR, we have progressed the automatic threshold detection method. Experiments show that this progressed algorism can provide satisfactory shot segmentation results in common video data. Secondly, how to apply classic key-frame extraction methods to compact video have been proposed in this paper.Thirdly, MPEG-7 descriptors and description schemes are detailed given. Then, feature definition, representation and retrieval forms in MPEG-7 as well as its application in video retrieval are discussed.At present, Most of research in content-based video retrieval is in the theory part and there is no successful system to all kinds of video. My paper mainly focused on the theoretical research and a universal video retrieval system is devised.Finally, the application prospect and future trend of CBR are approved in the summarization of the whole thesis.更多还原